The Open-Source AI Stack
RSS
All models

Models · qwen3-5

Qwen3.5 122B-A10B

Open Alibaba · 2026-02-16 · Apache-2.0

A mid-size Qwen3.5 mixture-of-experts: 122B total with about 10B active per token. The large footprint needs a high-capacity box, while the 10B active count keeps single-stream decode brisk; it is often run at IQ4_XS for long-context work.

Architecture

tokens in Embedding vocab not disclosed · qwen tokenizer × 62 layers Grouped-Query Attention RoPE + YaRN context 262,144 tokens MoE Router ? experts total · ? active per token Output projection tokens out
Schema-generated from data/models.yaml. Every label is auditable against the model's sources.

Specs

Architecture
moe
Total params
122B
Active params
10B
Context window
262K tokens
Attention
gqa
Position encoding
rope-yarn
Post-training
sft, rlhf
OSI-approved
yes
Data released
no
Training code
not released

Available quantizations

GGUF llama.cpp's container; the common local format, k-quants from Q2 to Q8. runs on llama.cpp, Ollama
AWQ Activation-aware 4-bit weight quantization for GPU serving. runs on vLLM, SGLang
GPTQ Post-training 4-bit weight quantization for GPU serving. runs on vLLM, SGLang, Transformers
MLX Apple MLX 4/8-bit layout for Apple silicon. runs on Apple MLX
FP8 8-bit float, frequently a native release on Hopper / Blackwell GPUs. runs on vLLM, SGLang, TensorRT-LLM

Verified via the Hugging Face model tree ↗. Community quantizations change over time; the families shown are those with published weights at audit time.

Notable innovations

  • · Sparse MoE with about 10B active per token
  • · Long native context

Sources